Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
FluSplat is introduced, a feed-forward framework for editing 3D scenes represented as Gaussian Splatting from sparse views, avoiding iterative test-time optimization. The method trains with a cross-view regularization scheme in the image domain, enforcing consistency through multi-view edits supervised with geometric alignment constraints. Results show competitive editing fidelity and significantly improved cross-view consistency with orders of magnitude faster inference compared to optimization-based approaches.
Forget slow, per-scene optimization: FluSplat edits 3D Gaussian Splattings from sparse views in a single feedforward pass, achieving cross-view consistency through a novel training scheme.
Recent advances in text-guided image editing and 3D Gaussian Splatting (3DGS) have enabled high-quality 3D scene manipulation. However, existing pipelines rely on iterative edit-and-fit optimization at test time, alternating between 2D diffusion editing and 3D reconstruction. This process is computationally expensive, scene-specific, and prone to cross-view inconsistencies. We propose a feed-forward framework for cross-view consistent 3D scene editing from sparse views. Instead of enforcing consistency through iterative 3D refinement, we introduce a cross-view regularization scheme in the image domain during training. By jointly supervising multi-view edits with geometric alignment constraints, our model produces view-consistent results without per-scene optimization at inference. The edited views are then lifted into 3D via a feedforward 3DGS model, yielding a coherent 3DGS representation in a single forward pass. Experiments demonstrate competitive editing fidelity and substantially improved cross-view consistency compared to optimization-based methods, while reducing inference time by orders of magnitude.